Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[mikeb]

Recently my line (talktalk 40/2) has had interleaving on the downstream where it used to be fastpath both ways (apart from a one off glitch a year ago, which I never really got to the bottom of). It's had interleaving of around 300 for a good couple of months now. The current stats from my Draytek 2760 are attached, does this give any indication of what's going on?

[Chrysalis]

ideally we need the uptime to try and calculate FEC and ES per day.

ECI hopefully is getting g.inp this spring so if that rollout goes ok your line may move to that.

[mikeb]

Uptime when those stats were taken was only about 17 hours (we had power outage all day yesterday).

Downstream FECS are now at 64969, so an increase of nearly 7000 in two and a half hours? All the other figures are about the same as they were.

[j0hn]

Interleaving was stuck on my line for about a year. The last 6 months of that I was ILQ green (below DLM limits) every single day, it just would not budge. I tried limiting my sync rate considerably lower than what I usually sync at and 2 days later my line changed to fastpath. Others have also been able to shake interleaving by capping their sync rate, reducing ES & FEC.

You need a Broadcom modem with telnet access to its CLI commands. A tutorial on how to do this is can be found here.

If you have the right kind of modem I would try setting a maxdatarate to 12000 2000 100000

It is my opinion that DLM looks at numbers of FEC when considering removing interleaving. Capping the sync can dramatically lower the FEC numbers. In my case FEC's were as high as 50,000 per min at normal sync rate, and down to 1000 an hour when interleaving gave way. ES were down to around 20 a day after capping sync.

Hope this helps

[mikeb]

Thanks John, I had wondered about capping sync speeds before, the modems try for a slightly too optimistic speed. I think it used to be pretty happy around 19mb, a couple of days ago I noticed that it was synced at 23mb down (with an attainable rate of 26mb), which I've never noticed that high before. Are both modems involved in the sync rate decision, or is it all commanded from the cabinet side?

Anyone know if it's possible to set sync caps on my Draytek router?

[BigJ]

I found a link to a pdf of telnet commands. No idea how upto date it is but I guess not. It mentions "adsl maxdnrate(for 2800 series)" but its for a different model series and ADSL only.

[mikeb]

I telnetted in and had a look. There is a command in there "vdsl snr" that can be used to set the snr margin. Interestingly it reported a current target margin of 5db (seems low, I've never seen it reported below 6), and a margin delta of zero. So I changed the margin delta to 30 (+3db, above 5, so aiming for 8). It's resynced at 16.5mb down with a margin of 9db. Interleaving is still on, but reduced to 242. I'll leave it like that for a while and see what happens.

[j0hn]

Excellent news you can set that with the Draytek.
Keep an eye on the ES numbers and FEC numbers.

[mikeb]

This is what the stats look like now, after 144 hours of uptime. What are reasonable figures for FECS?
I'm getting 1100 FECS per hour, is that a problem or normal?

Increasing the SNR margin has reduced the speed down to 16.5 Mb, but interleaving is still there, and above 200. Is the interleaving sticking around because of the number of FECS, or is that decided on other factors?

[burakkucat]

The number of FECs will not be a problem. Think of an FEC as a CRC which never occurred!

The DLM process takes a close interest in the number of ES that the circuit has accumulated . . . reduce the ES to zero, or close to zero, and the depth of interleaving will soon be reduced to 1.

[j0hn]

I can't make head nor tail of the Vigor stats I'm afraid.
What is 17593 "far end" ES. Is that upstream ES numbers? If so that's nearly 3000 a day!!

It is my opinion that DLM looks at numbers of FEC when considering removing interleaving.

It is known for definite that DLM uses ES numbers when making its decisions. Just with sticky interleaving being below what we believe to be the ES target is not enough to remove interleaving. To return to fastpath you need very low ES numbers, which usually coincides with much lower FEC numbers.

I personally think the FEC numbers matter under these circumstances. As we know ES numbers do count, you should focus on these. A period of around 10 days with sufficiently low enough ES should return the line to fastpath. For me this was below 20 ES each day, some days in the single figures.

My Upstream ES were 1 or 2 a day and are usually low enough they can be ignored. I don't understand your Vigor stats though. Surely those numbers can't be upstream ES.

[burakkucat]

I can't make head nor tail of the Vigor stats I'm afraid.
What is 17593 "far end" ES. Is that upstream ES numbers?

Me, likewise. 

There is one school of thought that uses the DS & US labels consistently relative to the CPE, even though the device being discussed/considered is the CO MSAN/DSLAM.

Then there is a second school of thought that uses DS & US labels strictly based on the device being discussed/considered.

Just to be different, Draytek use "far end" and "near end" labels . . . I presume we should associate such labels with the devices at each end of the link. As it is a CPE device producing the "report" I would assume that "near end" indicates the CPE and "far end" indicates the CO MSAN/DSLAM.

In the case of the ES, above, I am inclined to view it as the CO MSAN/DSLAM is reporting 17593 ES. But which "stream"? Which "direction"? Again, I am inclined to view it as DS relative to the CO MSAN/DSLAM . . . making it US relative to the CPE.

Using "transmit" and "receive" or even "go" and "return", relative to the device being considered, probably will not help. 

Now thoroughly confused, I think a simple solution would be not to use a Draytek device! 

[j0hn]

If those are actual ES numbers, it's kind of irrelevant if they are upstream or downstream. Interleaving isn't going to remove with numbers that high on either.

You can pick up a cheap Huawei HG612 (The original white OpenReach VDSL2 modem) which can easily be unlocked to obtain stats. If you really wish to "shake off" interleaving my recommending would be to source 1 of these. They can be found easily and cheap on a certain auction site. These allow the setting of MaxDataRate and simple continuous monitoring of the devices stats.

[mikeb]

The figures in the left column (downstream) reset to zero after my modem reboots. The right hand figures (upstream) never reset, I think someone said before that they are the stats from the cabinet end modem. I say never reset, maybe they reset if the cabinet looses power, or on some other event at the cabinet?

So on the downstream, I currently have 3 ES in the last 160 hours (so almost a week), which doesn't sound bad at all? The upstream ES of 17,149 is over a much longer period (possibly since my initial fttc commissioning date, which is about 15 months ago). The difference between the u/s ES figures in the two screen caps is 444, which equates to approx the last week as well. Is that excessive? Best case, assuming the total is over 15 months, that's about 40/day or 250/week, but obviously I've no idea what the spread is - it could have been perfect some days and terrible others, but it's definitely higher than downstream E/S.

What could be causing the differnce between u/s and d/s ES, presumably getting too close to the noise margin? And why am I ending up with interleaving being applied to the downstream while the upstream stays on fastpath? It looks to me like the cab modem is looking at the stats for the upstream side and deciding to apply interleaving to the downstream??     Is my modem or the cabinet modem to blame for this? It's almost like a basic compatibility issue?

I can control the SNR margin target on the d/s, but don't think I can influence the u/s snr?

I'd stuck with the Draytek over a HG612 modem to match the Lantiq chipset of my ECI cabinet. I thought that would be more beneficial than the extra stats logging available with the HG612. But maybe not?

[j0hn]

I would continue to monitor the line if you are certain about the ES numbers. They sound low enough for interleaving to drop off. If it doesn't you may need to limit the sync further.

It may be pointless though as your line may run with too many ES on fastpath to stay that way. There must be an additional noise source on your line since it used to run fastpath all the time. A crosstalker is a good example of this. My crosstalker causes hourly ES numbers to jump from under 100/hour to 400-500/hour. It's very obvious on my line as they turn their modem off every night.

The matching of Lantiq chipsets isn't always the most efficient. My ZyXEL VMG8924 with its Broadcom chipset performed better (both sync and ES numbers) than 3 different Lantiq chipsets on my ECI cabinet. The Huawei HG612 also performed better, though the ZyXEL outperformed it.